The article pairs the argument with a correlation view and a direct route to the PRNU lab.
Detection and attribution are different
A synthetic-media detector estimates whether patterns resemble a class of generated content. Source-camera attribution compares a questioned image with known images from a specific physical sensor.
A stronger question can be narrower
The same-camera hypothesis does not settle authenticity, but it can corroborate an acquisition story or reveal that the asserted source remains unsupported.
Limits are part of the result
Reference quality, image content, resizing, compression, crop geometry, and threshold selection all shape reliability and must remain visible.